Youth Aptitude as a Predictor of Adulthood Income

18 Pages·2015·0.22 MB·English

UUnnddeerrggrraadduuaattee EEccoonnoommiicc RReevviieeww
Volume 10 Issue 1 Article 9
2013
YYoouutthh AAppttiittuuddee aass aa PPrreeddiiccttoorr ooff AAdduulltthhoooodd IInnccoommee
Jaret Kanarek
Illinois Wesleyan University, [email protected]
Follow this and additional works at: https://digitalcommons.iwu.edu/uer
Part of the Labor Economics Commons
Recommended Citation
Kanarek, Jaret (2013) "Youth Aptitude as a Predictor of Adulthood Income," Undergraduate
Economic Review: Vol. 10 : Iss. 1 , Article 9.
Available at: https://digitalcommons.iwu.edu/uer/vol10/iss1/9
This Article is protected by copyright and/or related rights. It has been brought to you by Digital
Commons @ IWU with permission from the rights-holder(s). You are free to use this material in any
way that is permitted by the copyright and related rights legislation that applies to your use. For
other uses you need to obtain permission from the rights-holder(s) directly, unless additional rights
are indicated by a Creative Commons license in the record and/ or on the work itself. This material
has been accepted for inclusion by faculty at Illinois Wesleyan University. For more information,
please contact [email protected].
©Copyright is owned by the author of this document.
YYoouutthh AAppttiittuuddee aass aa PPrreeddiiccttoorr ooff AAdduulltthhoooodd IInnccoommee
AAbbssttrraacctt
I examine the relationship between youth aptitude and adulthood income. Using the National Longitudinal
Survey of the Youth 1979 cohort and OLS regression analysis, I test the hypothesis that a higher 1981
Armed Forces Qualifications Test (AFQT) score is directly related to a higher income in 2010, ceteris
paribus. First, a single regression equation is run for educational attainment subgroups at the time of
taking the AFQT. Second, a regression equation including total lifetime educational attainment, and one
that excludes it, are run to examine potential co-linearity between AFQT score and educational
attainment. The results show that AFQT is significant and positively related to adulthood income.
KKeeyywwoorrddss
AFQT, labor economics, human capital, youth
This article is available in Undergraduate Economic Review: https://digitalcommons.iwu.edu/uer/vol10/iss1/9
Kanarek: Youth Aptitude and Adulthood Income

I. Introduction

How important is youth aptitude in determining adulthood outcomes? This
question has been a central focus of many disciplines, including cognitive science,
psychology, and sociology (Knudsen, Heckman, Cameron, & Shonkoff, 2006).
More recently, economists have taken an interest in answering this question, as
much education and labor market policy has been based precisely on answers to
it. One such example is the Head Start program, which is a federally funded
comprehensive preschool education program provided to children of low-income
parents (DHHS). The underlying premise of Head Start is that hefty educational
investment during childhood will combat the effects of poor socioeconomic
factors on children’s economic future (DHHS). Although the efficacy of Head
Start is contentious (Barnett & Belfield, 2006; Deming, 2009; Garces, Thomas, &
Currie, 2002), similar programs such as the Abecedarian Project and the Perry
Preschool Program have been shown to be quite successful (Nores, Belfield, C.,
Barnett, & Schweinhart, 2005; Barnett & Belfield, 2006; Heckman, 2008).
Regardless, many economists have concluded that investment in youth education
projects like these is necessary for garnering improved labor market outcomes in
adulthood (Barnett & Belfield, 2006; Heckman, 2011; Heckman & Masterov,
2007; Knudsen et al., 2006). Some economists have gone so far as to argue that
such investment is key to national economic growth (Heckman, 2011; Knudsen et
al., 2006; MacEwan, 2013).

II. Theory

The framework underlying the content of this paper is the human capital theory,
which states that certain investments in an individual (e.g. education, physical
resources, parental resources, psychological support, etc.) should increase his
productivity and, resultantly, his earnings (Rosen, 2008). In this paper, the
cumulative human capital skills attainment at a point in time will be regarded as
“aptitude.” In context of “youth aptitude,” human capital theory is highly attuned
to the relationship between youth aptitude and adulthood outcomes, i.e. earnings.
For one, human capital is an economic resource that, like any other, is
scarce—time, money, and the amount of skills an individual can learn and apply
are all limited—and requires efficient allocation. Thus, the questions of how
much and as well as what time human capital should be invested are germane to
the theory (Rosen, 2008). Whether human capital should be invested early in
childhood, in adulthood, or sometime in between is a central question in guiding
many economic policies.
Additionally, human capital theory includes economic tools that provide a
clear and precise groundwork for hypothesis formulation and testing. The age-
Published by Digital Commons @ IWU, 2013 1
Undergraduate Economic Review, Vol. 10 [2013], Iss. 1, Art. 9

earning profile is one such tool. The age-earning profile theoretically has an
inverse “U” shape, whereas; as age increases, earnings rise until a point of peak
earnings, after which earnings begin to decline. All things equal, higher human
capital levels should garner a higher age-earning curve (Rosen, 2008).
Consequently, if pre-market entry youth aptitude is higher for one individual than
another, then all things being equal, the individual with a higher aptitude should
be on a higher curve upon entry into the labor market. As such, higher earnings
would be represented for the higher level of human capital attainment.
The theory also offers theoretical explanations of youth aptitude as it
relates to adulthood income. Since youth aptitude is indicative of an individual’s
capacity to internalize human capital investment, a youth with a high aptitude will
be able to acquire more human capital than a youth with a lesser aptitude in the
same amount of time. Accordingly, the youth with a high aptitude should also see
higher earnings given his greater level of human capital.

III. Literature Review

There are other components of human capital theory as it applies to the
relationship between youth aptitude level and adulthood earnings. I will
specifically examine three of these components in the following literature review,
which will then lead me to a formal presentation of my hypothesis. The three
components are as follows: 1) the factors which most determine youth aptitude, 2)
the measures of aptitude, and 3) the relationship between youth aptitude, earnings,
and labor market performance.
There are myriad human capital factors, both direct and indirect, that
influence youth aptitude. Direct investments such as education play a critical role
in the development of youth aptitude levels (Barnett & Belfield, 2006; Cascio &
Lewis, 2006; Caspi, Wright, Moffitt, & Silva, 1998; Carvalho, 2012;
Frederikkson & Öckert, 2005; Griliches & Mason, 1972; Hansen, Heckman, &
Mullen, 2004; Hause, 1972; Heckman, 2008; Heckman, 2011; Heckman &
Masterov, 2007; Knudsen et al., 2006; MacEwan, 2013). The innate ability of a
child to internalize human investments may also directly influence aptitude levels.
A child with higher innate ability is, theoretically, able to better convert human
capital investments into greater returns than someone with a lesser innate ability
to do so. Not surprisingly, IQ scores and general tests of cognitive levels and this
innate ability have been shown to be indicative of productivity and thus earnings
(Carvalho, 2012; Caspi et al., 1998; Heckman, 2008; Jones 2005; Jones, 2011).
Similarly, Currie (2009) and Carvalho (2012) found that there was a strong
positive link between a child’s health and his adulthood education and income
attainment.
https://digitalcommons.iwu.edu/uer/vol10/iss1/9 2
Kanarek: Youth Aptitude and Adulthood Income

Indirect human capital investments are those that either help or hinder an
individual’s attainment of human capital but are not investment in human capital
per se. These include socioeconomic, geographic, and familial situations and
many other factors. Family background variables, such as family poverty status,
the educational attainment of a child’s parents, the size and makeup of the family
unit, among others, strongly determine aptitude levels (Caspi et al., 1998; Currie,
2009; Currie & Thomas, 1999; Heckman, 2011; Israel & Seeborg, 1998; Knudsen
et al., 2006). Guo (1998) even argues that the time in which youth experience
poverty plays a significant role in aptitude. These factors are entirely external, yet
there is also significant evidence on the role of personal factors as well. Heckman,
Stixrud, and Urzua (2006) found that non-cognitive factors (e.g. self-esteem,
personal expectations) are strongly related to aptitude and labor market
performance in that they strongly influence schooling, occupation, and behavioral
decisions. Carvalho (2012) and Heckman (2008) seconded these findings in their
own research, although Carvalho defined non-cognitive factors more broadly than
Heckman. Congruently, Israel and Seeborg (1998) found that educational
expectations are strongly related to actual educational attainment, and as such,
aptitude. Caspi et al. (1998) also found that anti-social and other behavioral issues
strongly correlated with decreased aptitude and labor market outcomes.
Evidently, there are myriad human capital factors, both direct and indirect,
that most shape youth aptitude. For this reason, there is a large body of literature
dedicated to the measurement of this aptitude. IQ and other modified IQ tests
have become popular measures of human capital (Jones 20011; Jones 2005).
Further, many of the same factors that influence aptitude also directly influence
IQ, such as socioeconomic, environmental, and familial factors (Armor, 2003).
However, there has been little focus given to how IQ, in turn, can be used to
measure these factors. As far as the literature is concerned, IQ is only suitable for
measuring cognitive skills strongly associated with human capital (Jones 2005).
As such, the explanatory power of IQ as a measurement of human capital
attainment as a whole, and therefore aptitude, is limited.
One other prominent measure of aptitude is the Armed Forced
Qualifications Test (AFQT). The AFQT is a component of the Armed Services
Vocational Aptitude Battery (ASVAB), which is used by the U.S. military to
determine enlistment qualification. Unlike IQ, the AFQT has been thoroughly
studied to determine exactly what aspects of aptitude it measures. Yet, the
conclusions on this issue have varied significantly. Herrnstein and Murray (1994)
notoriously posited that AFQT as a measure of innate and heritable intelligence in
The Bell Curve. Conversely, many studies refute this claim and conclude that
AFQT is more a measure of socioeconomic background than innate intelligence
(Caspi et al., 1998; Currie, 2009; Currie & Thomas, 1999; Guzman, 2001; Israel
& Seeborg, 1998). Other economists have instead concluded that AFQT is a
Published by Digital Commons @ IWU, 2013 3
Undergraduate Economic Review, Vol. 10 [2013], Iss. 1, Art. 9

function of educational attainment (Cascio & Lewis, 2006; Goldberger & Manski,
1995; Griliches & Mason, 1972; Guzman, 2001; Hansen, Heckman, & Mullen,
2003; Hause, 1972; Munday, 2001). Such a narrative bodes well with the human
capital theory’s emphasis on educational attainment as it relates to future
earnings. Indeed, it is in these studies that the strongest relationship between
AFQT scores and any other factor has been established. Johnson and Neal (1996)
prominently published findings that AFQT accurately measures differences in
human capital-based skills, although they do conclude that those differences can
ultimately be traced back to youth socioeconomic differences. Heckman, Stixrud,
& Urzua (2006) conclude that non-cognitive factors play a significant role in
determining AFQT scores.
In summation, it is inconclusive as to which exact human capital factors
most determine aptitude as measured by the AFQT. What is conclusive from the
literature is that AFQT is generally a measure of aptitude, although which factors
contribute most to this measured aptitude is unclear and deserves to be the focus
of future research. For the purposes of this paper, then, I will operate under the
assumption that AFQT represents general youth aptitude. I will do so having
given full consideration to the critics of this assumption, the AFQT, and human
capital theory generally (Rogers III & Spriggs, 1996). Additionally, it is important
to note that much of the literature concerning the AFQT has used the National
Longitudinal Survey of Youth (NLSY) 1979 cohort (Cascio & Lewis, 2006;
Currie & Thomas, 2009; Guzman, 2001; Hansen, Heckman, & Mullen, 2004;
Heckman, Stixrud, and Urzua, 2006; Johnson & Neal, 1996; Munday, 2001).
The last part of my literature review will focus on the relationship between
youth aptitude (as measured by AFQT), earnings, and labor market performance.
Generally, there has been a positive link found between youth aptitude and
adulthood earnings in the literature (Caspi et al. 1998; Heckman, 2008; Heckman
& Masterov, 2007; Heckman, Stixrud, & Urzua, 2006; Johnson and Neal, 1998).
This has been an inherent part of these past discussions given that youth aptitude,
as measured by AFQT, represents human capital attainment and human capital
attainment is, in theory, positively related to earnings and labor market
performance. Interestingly, the evidence on whether youth AFQT scores are an
accurate predictor of adulthood income is minimal. Little focus has been given to
this relationship relative to the amount of literature examining the AFQT, and the
evidence that does exist indicates that youth AFQT scores are not an accurate
predictor of adulthood income (Griliches, 1972; Hause, 1972). Largely, this is a
result of the AFQT and educational attainment being utilized as two distinct
measures of aptitude within the same regression, despite the fact that the latter
plays a large part in the former (Cascio & Lewis, 2006; Griliches & Mason, 1972;
Hansen, Heckman, & Mullen, 2004; Hause 1972).
https://digitalcommons.iwu.edu/uer/vol10/iss1/9 4
Kanarek: Youth Aptitude and Adulthood Income

My research will be distinct from the existing literature in two ways. First,
it will regard AFQT as a general measure or sum of aptitude, and not as a result of
one or two overriding human capital factors such as education, socioeconomic
background, or innate ability. Second, it will take a unique approach to isolating
the effects of AFQT apart from educational attainment. Although there is
expected co-linearity between the AFQT score and education, my empirical
model will be structured so that this co-linearity can be both controlled for and
identified.

IV. Hypothesis

The literature suggests that then there should be at least a marginally positive
relationship between youth AFQT scores and adulthood income. Human capital
theory suggests that higher youth AFQT scores should be strongly related to
higher adulthood income. These relationships should be accurate under the
assumptions that the AFQT is an overall measure of youth aptitude, and aptitude
represents cumulative human capital accumulation measured at a point in time.
As such, the hypothesis I will test is: A higher 1981 AFQT composite
score is directly related to a higher income in 2010, ceteris paribus.

V. Data

In accord with much of the prior literature, I will be using the National
Longitudinal Survey of Youth (NLSY) to test my hypothesis that a higher 1981
AFQT composite score is directly related to a higher income in 2010, ceteris
paribus. The NSLY is a panel dataset, which surveys a cohort of individuals over
time. Specifically, I will be studying the 1979 cohort, which interviewed 12,686
individuals between the ages 14 to 22. Due to the specific variables I will be
examining, as well as cohort dropouts, my sample size will be limited to 6,671 of
the individuals in the cohort. Of this group, 48.3 percent are male, 50.3 percent
are racial minorities, and in 2010 77.8 percent resided in urban areas.
Additionally, 90.2 percent had completed between 9th and 12th grade of formal
education by 1981, the time at which the AFQT was universally administered to
the cohort. The survey of this cohort began in 1979, and re-interviewed the cohort
nearly every year until 2010; the most recent year data is available.
The NLSY 1979 cohort offers distinct advantages. First, the fact that it is a
panel data set makes it possible to chronicle and study the long-term effects of
lifetime experiences. Second, it includes many of the necessary variables for this
study, including AFQT score and income. Third, it surveyed youth, which make
the NLSY sample perfect for studying youth aptitude as it relates to adulthood
income. Fourth, the depth of variables included in the NSLY provides for related
Published by Digital Commons @ IWU, 2013 5
Undergraduate Economic Review, Vol. 10 [2013], Iss. 1, Art. 9

future research and comparative analysis of findings. One notable drawback of the
dataset, however, is that it does not measure school quality, which is particularly
challenging for human capital theory-based research problems. This can result in
a potential skew in the data and results, as both years of schooling and school
quality are important measures of human capital attainment.

VI. Empirical Model

Regression analysis will be the heart of my empirical model. In this section, I will
first present my regression equation and describe its components. Second, I will
identify and explain the two general capacities in which I will run this regression.
The first way will involve running the regression for different subset samples.
Subset samples will be distinguished by years of schooling at the time of taking
the AFQT. The second way will involve running two regression equations for the
whole sample. The first regression equation will include total years of schooling
as an independent variable, and the second will exclude total years of schooling.

A. The Regression Equation

My regression equation, and the subsequent description of its components
(in Table 1), is as follows:

LnWage = α + β (AFQT1981) + β (YearsEduc2010) + β (Urban) + β (Age2010)
1 1 2 3 4
+ β (Male) + β (Minority) + ε.
5 6 i

Table 1: The Variables of The Regression Equation
Variable Description Expected Sign
LnWage Independent variable; the natural log of N/A
2010 wages; continuous
AFQT1981 AFQT composite score; Continuous Positive
YearsEduc2010 Years of formal schooling by 2010; Positive
Continuous
Urban Correspondent’s geographic setting in Positive
2010; Urban = 1, Rural = 0
Age2010 Correspondent’s age in 2010; Positive
Continuous
Male Gender of correspondent; Male = 1, Positive
Female = 0
Minority Minority if African American or Negative
Hispanic, non-minority if white;
https://digitalcommons.iwu.edu/uer/vol10/iss1/9 6
Kanarek: Youth Aptitude and Adulthood Income

Minority = 1; Non-minority = 0

This regression equation is set up to measure the effects of AFQT1981 on
LnWage, the main independent variable and the dependent variable, respectively.
Specifically, I will be focused on the sign and magnitude of β , which is the
1
coefficient of AFQT1981. A positive sign for β indicates that each additional
1
score higher for AFQT1981 increases LnWage. The magnitude of β indicates
1
how much this increase is. As such, a positive and large β should indicate that
1
each additional score higher for AFQT1981 should result in large increase in
LnWage. Of less concern are my human capital control variables, including
YearsEduc2010, Urban, Age2010, Male, and Minority. YearsEduc2010 is
intended to control for total educational attainment, which human capital theory
suggests should be positively related to LnWage. Another weakness of NLSY is
that it does not measure neighborhood effects on income. Albeit, those that live in
urban areas tend to have higher incomes and more resources available to them, as
compared to rural areas. As such, Urban is included to control for socioeconomic
surroundings to an extent. Age2010 is included to control for the age of the
cohorts, as age is positively related to income in human capital theory. Male and
Minority are included as controls for gender and race-based wage differentials,
respectively.

B. Regression Analysis for Years of Schooling Subsets

The first way in which I will run the regression equation identified in section
IV.A will involve separate regression analysis for different subset samples. Subset
samples will be distinguished by years of formal schooling at the time of taking
the AFQT (YearsEduc1981). YearsEduc1981 is continuous and ranges between
1st grade and an 8th year of college. Whereas, YearsEduc1981 = 8 represents the
completion of 8th grade as the highest level of formal schooling at the time of
taking the AFQT. As shown in Table 2, I will be running separate regressions
only for YearsEduc1981 = 9, 10, 11, 12, 13, and 14. Subsets YearsEduc1981 = 9,
13, and 14 have much smaller sample sizes and are unlikely to include any
statistically significant variables.

Table 2: 1981 Years of Schooling Subset Sample Sizes
YearsEduc1981
9 10 11 12 13 14
Sample Size 514 1055 1002 2491 562 389

The rationale for running separate regressions for each subset is to control
for educational attainment level in the resultant AFQT scores. An additional year
Published by Digital Commons @ IWU, 2013 7
Undergraduate Economic Review, Vol. 10 [2013], Iss. 1, Art. 9

of education is strongly related to a higher AFQT score (Cascio & Lewis, 2006;
Goldberger & Manski, 1995; Griliches & Mason, 1972; Guzman, 2001; Hansen,
Heckman, & Mullen, 2003; Hause, 1972; Munday, 2001). Thus, running separate
regressions for differentiated YearsEduc1981 helps to provide a more accurate
representation of the relationship between youth aptitude, as measured by AFQT
scores, and adulthood income. Further, it is well established in the human capital
literature that education plays a large roll income level (Card, 1998). In short,
these sample subsets are a unique control for educational attainment. The
downside to this approach is that it further limits the size of my samples, and thus
makes the attainment of statistically significant variables more challenging. This
concern, however, will be addressed with the first regression in Section IV.C,
which will not be restricted to any specific YearsEduc1981 subgroup.
Despite the use of separate regressions for each subgroup, the purpose of
each regression remains the same, i.e. to measure the effects of AFQT1981 on
LnWage. The sign, magnitude, and statistical significance of β , the coefficient of
1
AFQT1981, will still be my main focus in the analysis portion of this paper.
Because I am controlling for YearsEduc1981, each regression should theoretically
yield similar results.

C. Regression Analysis With and Without Control for YearsEduc2010

This will include two regression equations, both of which will be run for the
entire sample. The first regression equation will be the same as the one used in
IV.A and B. This will provide an overall picture of the relationship between
AFQT1981 and LnWage. The second regression will be modified to exclude the
control for YearsEduc2010. As such, this second regression will be as follows:

LnWage = α + β (AFQT1981) + β (Urban) + β (Age2010) + β (Male) +
1 1 2 3 4
β (Minority) + ε.
5 i

The expectation is that the second equation will yield a significantly
higher positive coefficient for AFQT1981 when not controlling for
YearsEduc2010. The rationale for using this second equation is to examine the
co-linearity between these two variables. If AFQT1981 is truly a measure of
aptitude, i.e. accumulated human capital attainment at the time of the test, then a
higher AFQT score for an individual is also likely to be predictive of higher future
educational attainment. In turn, YearsEduc2010 includes the years of education
that contributed to AFQT1981. Thus, if β rises by a large magnitude, then the
1
data suggests that the co-linearity between the two variables is quite large. This
would further support the choice to use human capital theory as the primary
framework for analysis in this paper.
https://digitalcommons.iwu.edu/uer/vol10/iss1/9 8